Kneller Lecture, Comparative & International Education Society, Annual Conference, New Orleans, 2013 (preliminary version). Published in Prospects. June. 2013. Starting in the 1970s, Brazil developed a very complex system of quantitative assessments of education at all levels, making extensive use of statistical information and tests, in an effort to improve and maintain the quality of its education. However, with some exceptions, particularly at the graduate level, Brazilian education standards remain low, with few signs of improvement. After reviewing the existing assessment systems and their evolution, this Viewpoint argues that, although these assessments are a necessary component of any successful policy for educational improvement, they may fail without the appropriate technical and institutional considerations. Technical considerations include the need to avoid situations of reification, when the indicators, rather than education itself, become the main goal to be pursued; when obscure statistical estimations replace well-grounded psychometric measures aligned with existing curricula; when low-stakes tests are used for high-stakes purposes; and when the pedagogical, psychometric, and statistical procedures are not open to regular outside peer review. Institutional considerations include the need to make the assessment agency independent from government and other stakeholders, and to consider the large differences across the country’s regions and populations.